Title: Current status of Pipeline process for binary coalescence search in TAMA300
1Current status of Pipeline process for binary
coalescence search in TAMA300
- GWDAW2002
- _at_ International Institute for Advanced Studies
- 18 December 2002
Daisuke TATSUMI National Astronomical Observatory
of Japan
2Overview of pipe-line process
- Detector output is recorded by Data Acquisition
System as a raw data. - Raw data is transferred to suitable format for
each GW search by pre-process. - And then, pre-processed data is distributed to
the TAMA collaborators. - Finally collaborator makes analysis of
interesting GW signals.
? Data Acquisition
? Pre-process for GW search
? Network Data Distribution
? GW signal search
31. TAMA Raw Data
- High Speed Data Acquisition (HDAQ)
- Sampling Clock 20 kHz
- Number of Data in a Frame 216 65536
- Number of Frames in a File 20 frames / file
- File Length 65.536 sec
- Online Calibration
- Continuously monitoring a Detector Response
- at a fixed frequency of 625 Hz by exciting
mirrors - Time Information
- IRIG formatted waveform from GPS receiver
- is recorded together with Detector output
signals. - By using this signal, we can obtain an accuracy
of - 1 micro second.
42. Pre-Process
- Data Quality Checking (ADC over-range, lock
status) - Extraction of Time Information from IRIG signal
- Online Calibration
- Noise Spectrum Density in Strain
- Averaged Power Spectrum
- Specification of Searching Grid
Matched filter analysis
5Data Length for Inspiral Search
- Change data length to be suitable for Inspiral
search - The above temporary data of 16 frames with one
frame-overlapping - are analyzed. Details of it are described in
the following section. - And then, these analyzed information is
recorded in frame format - as a Proc Data.
20 frames
20 frames
20 frames
Raw Data
20 frames 65.536 sec
Temporary Data
16 frames
16 frames
16 frames
16 frames
16 frames with 1 frame-overlapping
1 frame 52.4288 sec
Proc Data
1 frame
No ADC data is included in.
6Moving Average of Noise Power Spectrum
Because the Detector noise is not stable, we
should evaluate averaged noise power spectrum
time by time. But simple definition of it causes
the big influence of burst like noise. So we use
inversed power spectrum in moving average. This
method presents very good performance like this.
1/
73. Network Data Distribution
Data Transfer gets the great benefits of high
speed network for academic researcher called
Super SINET. It has been constructed as National
Project and began to operate in January of
2002. Since this summer, we can distribute
TAMA data of 1GB/hour continuously.
This is supported by computer center of each
institute.
8Task List
Distributed data was analyzed as follows. The
details of each analysis will be given in the
following sessions.
- Inspiral Search
- NAO Tatsumi
- Osaka U. Takahashi / Tagoshi
- (session8-1, 2)
- Continuous
- U. Tokyo Soida (session6-1)
- Burst
- U. Tokyo Ando (session7-1)
- Ring Down
- NAO / Osaka-C.U Tsunesada / Kanda
9Elapsed Time for 20 MB (1 min.) Data Transfer
For 1 min. data transfer, 5 seconds is used
by rcp. In the case of local copy, time is 2
seconds and secured copy spends about 20 seconds.
- ltcpgt (local machine) 2 sec
- ltrcpgt
- within laboratory 2 sec
- to Osaka-U 5 sec
- ltscpgt
- within laboratory 18 sec
- to Osaka-U 17 sec
To keep network security together with high
transfer rate, rcp with access restriction and
VPN technique are used.
104. Inspiral Search
- Necessary Computing Power is depend on number of
templates to search in the mass region.
With current computing power of 16 PC
cluster, more than 1 solar mass region can be
searched on time. On time means that analysis
time is shorter than data length.
11Results of DT7 data (July 2002)
This figure shows event distribution of a result
combined with matched filtering and chi-square
testing. In the next international coincidental
observation, this Inspiral search result will be
available on time around neutron star mass region.
Preliminary Result
12Summary
Pipeline Process is under construction for
Online GW search. Especially for
Inspiraling Neutron Star Binaries, a search
result will be available on time in the next
international coincidental observation. In
the future, this kind of pipeline process is
applied for Detector Diagnosis and Continuous
Signal Search.